I like the concept, but I feel that your algorithm for generating an initial summary needs a lot of work. For example, I was experimenting with the API:

The input page:

http://bookflavor.com/dukan-diet-2-steps-lose-weight-2-steps...

The API results:

http://gistpoint.com/get?u=http://bookflavor.com/dukan-diet-...

Rather than choosing a snippet of the main text, it chose a few words from one of the menu items...

If you can fine tune the algorithmic generation of snippets in the API this could be very useful. Keep up the good work!